Crypto Perpetual Pairs Trading with Correlation Screening
Summary
This tutorial outlines a statistical arbitrage approach for crypto perpetual contracts. It screens historical closing prices for highly correlated pairs, then trades deviations in their price ratio: when the ratio is above its reference level, it buys one contract and shorts the other; when below, it reverses the legs. The reference ratio updates over time. Python examples cover market data collection, correlation analysis, a simple exchange simulator, and a backtest using hourly candles. The article reports that several selected pairs had very high measured correlations and says four pair tests looked favorable, though it gives limited performance detail.
The author acknowledges a central bias: the initial pair selection used future data, so a split sample produced weaker results. High correlation does not guarantee stable mean reversion. Relationships can change, dislocations can widen, thin liquidity and fees can erode returns, and extreme markets can create losses. Suggested improvements include periodic pair reselection, stop rules, and diversifying across pairs.
Key ideas
- The strategy trades two related perpetual contracts when their relative price departs from a moving reference ratio.
- Historical price correlation is used to screen candidate pairs, with Pearson correlation shown as the measure.
- The tutorial supplies a basic simulator and hourly backtest, but its first pair selection uses future data.
- A split-sample exercise weakens the results, highlighting selection bias and limited evidence.
- Changing correlations, widening deviations, execution costs, and liquidity can undermine the strategy.
Tags
Cited by
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.